Optic flow density modulates corner-cutting behavior in a virtual reality driving task

When driving on a roadway, the visual guidance of steering is thought to rely significantly on optic flow (OF). It is perhaps surprising that patients with cortically-induced blindness (CB) over a visual quadrant or hemifield demonstrate greater biases in lane keeping than visually-intact controls (Bowers et al. 2010) because OF is spatially correlated and heading judgements can be accurate using only a portion of the flow field (Warren and Kurtz 1992). One possible explanation for these biases is that the “blind” region acts as a source of signal-dependent noise that biases perception of heading during steering. In this preliminary study, we asked to what extent OF signal strength impacts steering accuracy. Seven visually-intact subjects immersed in a simulated environment seen through a Vive Pro were tasked with staying within a procedurally-generated roadway while traveling at 26.6 m/s. Between turns, we varied turn direction (right/left), turn radius (35, 55, or 75 m), and the density of texture elements that provide OF information (low, medium, high). Results revealed that average signed divergence from road center increased with OF density (low OF: 0.13 meters ±0.25 (SD), medium: 0.38±0.28, high: 0.55±0.34), where positive values indicate bias towards the road’s inner edge (i.e. corner-cutting). Our findings suggest that lane biases and corner-cutting behavior are strongly impacted by texture density (and thus OF) in the virtual environment, which is similar to the effect of global flow rate found by Kountouriotis et al. (2016). However, additional work is needed to specifically attribute the behavior to changes in perceived speed or changes in the relative weighting of OF vs. alternative information. Future work will be extended to include CB patients and test whether this effect of OF on steering behavior is mitigated by characteristics of the “blind” field.